Thermal conductive epoxy: powdered diamond, only $1200 for 5g

Seems a little high priced, to me. Abrasive graded grit in diamond used to be up to $40/gram (last time I bought any). Epoxy, was even cheaper.

Or just buy 'em. Ten-millimeter round faceted will set you back maybe 2 bucks each, after shipping

Find a cheaper source. For 10 mm round they want $135/ea:

However, you can get 10 mm ruby laser rods from Russia at reasonable prices. Just slice, cut, and polish your own gems:

The thing about particle-filled epoxy is the the particles increase the gap, so increase thermal resistance. The particles are still mostly separated by epoxy, a terrible thermal conductor.

Dow 340 silicone grease has very fine fill particles, and squishes down somewhere below 100 micro-inches thickness with moderate pressure. I bet that has lower theta than diamond-grit-filled epoxy, given flat surfaces.

The price differential must be 1e4 or so.

Cosmetic grade Boron Nitride is pretty cheap and very fine. Plus, it'll hide your wrinkles.

The best particles are no particles. But if the surfaces aren't super-flat, fill is needed, so something with very small particles will help. It's important that it flow out of the contact areas into the gaps.

Given a typical power mosfet (very flat) and a typical extruded heat sink (not very flat) Dow 340 will cut theta by a factor of 3 or so. I'm guessing (pat pending) that diamond-grit goop wouldn't be dramatically better.

There's multiple processes and recipes; I was looking at a less expensive variant:

The problem with grease is that it tends to pump itself out from under the hot spots under temperature cycling. If one could get the diamond particles to embed themselves in both sides of the metal, they'd stay put and probably improve the theta.

Cheers

Phil HObbs

I think that's correct - the grease would have to be very heavily loaded and under extremely high clamping force to embed the particles and make a dramatic difference.

Most of the greases and thermal (not electrically conductive) adhesives are filled with hexagonal Boron Nitride and/or Aluminum Nitride. Both can be had in powder form - micron size or less. BN is very soft like talc or graphite (sometimes sold as white graphite for its lubricating properties) so particles could not produce much of a gap. AlN is harder but milled fine enough it might be slightly better than BN.

The silver and copper filled stuff makes me nervous if it's in grease form - I wouldn't want that crap dribbling into the electronics.

I tried toothpaste a while back on a 125W AMD CPU, as I couldn't find my stuff and wanted to get it done. It ran all of 1C hotter than with zinc oxide paste. That's silica IIRC.

NT

That's the same supplier and page that I linked, except I'm on Page 2, where the 10 mm prices are listed. $135/ea.

I couldn't find anything on making rubies in a microwave oven, but did find this on making zircons: and diamonds: in a microwave oven. Do you perhaps have a better reference for rubies?

In a past life, I was responsible for getting the heat out of RF power transistors. I posted my results, which didn't quite agree with the internet published orthodoxy, about 2 years ago in this newsgroup. Now, I can't find it. Argh.

Very roughly, I got best results when I polished both the aluminum heat sink and the base of the power xsistor. The idea was to get as much metal to metal contact as possible. What worked best was no grease, no powders, and no mixes of grease and powders. Just metal to metal. However, that requires that both the heat sink and xsistor be flat and parallel, not twisted, bent, or torqued. Measuring the flatness of the heat sink was fairly easy. I can't say the same for the xsistor. So, I just tested for stiction against a known flat gauge block. Good enough.

During the testing, I tried various powdered metals without any greasy binder. I didn't have a wide selection of powders available, but I managed to try silicon carbide, graphite, metallic silver, alumina, and copper. I would say that powdered silver worked about the same as Arctic Silver heat sink goo. The best of the bunch was silver. W/m*K Diamond 1000 h-BN 600 c-BN 740 Silver 406 Copper 385 Gold 314 Aluminum 205 Graphite 200 Carbon 150 SiC 120 Brass 109

The idea behind the powder was to force the powdered metal particles into the aluminum heat sink and copper xsistor base, filling in the cracks and crevasses in the metal. In order to make that work, I had to clean/degrease the parts, smear on the powder, wipe off the excess, attach the transistor, and compress the mounting flange. If I had left a large quantity of powder on the heat sink, it would make things much worse. Inspecting the surface with a microscope showed that most (not all) of the particles were imbedded in the aluminum.

However, the copper xsistor base was another story. Most of the particles would just stick to the surface and not fill the cracks. I could beat on the xsistor to force the particles into the xsistor base, but that would cause the base to loose its flat surface. I could have left the gold plating on the base, but then it wouldn't be flat.

One thing I didn't try was gold leaf. Working with power transistor size pieces of gold leaf is difficult. It sticks to literally anything and will not let go. It also likes to wrinkle creating lumps. However, with a fairly high thermal conductivity and the ability to fill in cracks, it's still tempted to fight the problems.

The bottom line is that powder only made sense if both metals were soft, polished, flat and parallel. Controlling the amount of powder was difficult because the amount needed depended on the type and depth of the surface features, cracks and roughness.

In the end, we used a polished heat sink and xistor for the really high power stuff, and conventional heat sink goo for most everything else. It was possible to get an improvement by using some of the aforementioned methods, but it just wasn't worth the time and effort.

What, you didn't try toothpaste? :)

NT

If I repeated the tests today, I wouldn't bother with toothpaste. Toothpaste is 20-42% water, which has a lousy thermal conductivity (0.6 W/m-K). If I evaporated the water AFTER installing, there will be air gaps. No thanks.

Instead, I would try gold leaf (314 W/m-k) again and add flake graphite (750 W/m-K): "THERMAL CONDUCTIVITY OF GRAPHITE FLAKE COMPOSITES"

Also, note that the differences between the various compounds is in the area of 0.1 C/Watt. That doesn't seem like much until you try to dissipate 150 watts per device: 0.1 * 150 = 15 C which can be the difference between running hot and meltdown.

Do you think so? That would only be a problem if air moved in, to replace the grease in the gaps.

hence the smiley

how would gold leaf help? To get it to squash into the crevices and out of the both-flat areas you'd need to put huge force on it. If you don't, it's just another impedance in the way. Unfortunately the same is true of any so lid - and also I suspect why grease didn't work well for you. As the gap na rrows, grease at some point becomes too stiff to flow through such tiny gap s, so you end up with grease separating the 2 surfaces.

NT

I don't do smileys. Hieroglyphics went away with the ancient Egyptians.

Thermal resistance, not impedance. There's no imaginary part.

Gold leaf has better thermal conductivity than most of the pastes and powders. Visualize to non-mating surfaces full of cracks, but reasonably flat. The surfaces will touch only at the highest peaks, leaving air pockets in the gaps and cracks. The gold leaf is thin enough that the touching peaks will push aside some of the old and either down the sides of the peaks, or into the cracks. It won't fill the cracks as a grease or powder might, but will have three high thermally conductive materials touching. Unfortunately, I don't have a sectioning saw necessary to produce a microscope cross section photo to see what the gold is really doing.

Nope. The problem with grease is determining how much to use. If the grease is thick, and I use too much, the cracks are filled, but there's also a rather thick layer of comparatively high thermal resistance grease sitting between the heat sink and the xsistor. If I use very thin grease (i.e. oil), then it just leaks out of the joint and loosens the torque on the mounting screws. A compromise is needed.

The best way to do grease, paste, and powder is to thinly smear the paste onto both the aluminum and copper down to where you can see both grease and metal sticking through the grease. In other words, fill the cracks, but don't go over the metal peaks. Then assemble and compress. That uses an amazingly small amount of thermal grease. However, there are a few problems. It doesn't work well if either part is twisted. If not done right, it will also leave the sides of the metal peaks uncovered. I originally used the end of an acrylic ruler to apply the grease. That worked well until the end of the ruler showed grooves resulting in grease ridges on the heat sink. I'm not sure what to use now, but I suspect it will need to be harder than aluminum.

Anyway, my problems with grease were probably from applying the grease either too thick or too thin.

Dow 340 squishes out from under a part with modest pressure. It flows. I don't think that too thick is a problem.

I measured 100 microinches thickness added by some, which was my limit of measurement resolution with a micrometer. It may squish thinner.

Bubbles/air gaps could be a problem. I think that the best application is one drop in the middle of a part, then press onto the sink with a circular motion, until grease exudes all around the edges of the part.

You reckon that with only hand pressure, the solid metal gold will flow sid eways from under the peaks. I have to say that sounds optimistic - you've c learly done a lot more work on this than I have, but that is a bold claim, and I think would warrant some sort of measurement of whether that does hap pen.

no, and there's its other problem. There's so little bulk that I wouldn't e xpect it to do much in the crevices.

I wonder if there's any other way to do it

FWIW my observations of used parts in commercial domestic equipment is that there's always a nearly whole layer of grease left between the 2 surfaces. Maybe that's just because they're so far from flat, I don't know.

Would a single scrape with a regular steel scraper produce significant surf ace damage? If so I guess you just have to keep regrinding something soft.

I came across the flow problem when trying to pass regular engine oil throu gh a 1mm nozzle. Getting it to move at all proved beyond the physical limit s of the equipment.

NT

Some friends of mine developed a couple of interesting technologies for metallic thermal interface materials (TIMs). One is an indium-gallium liquid alloy that was used in some higher end Apple machines. It works the best of anything, but is vulnerable to corrosion. The other is indium foil with a very small quilt pattern embossed in it. Both work dramatically better than the best paste. The quilt pattern allows the metal to conform better to the surfaces, because it only has to flow a short distance to get out from under a high spot. Either one is good enough for a 2500-sun solar concentrator--in fact it was the metal TIM that made the concentrator possible. (I did a little work on the solar project, but not on the TIMs, except to cheer from the sidelines.)

The best paste as of 2007ish came in at about 3.5 W/m/K. It was made by a Japanese outfit whose name I forget--it had tiny flat metal flakes in it, which were supposed to form stacks so that most of the heat went through a solid metal path. It turned out that the stacks performed worst under temperature cycling. Since the stacks were pinned in place, expansion of the grease washed the small particles out from between the flakes, which dramatically increased the thermal resistance.

Cheers

Phil Hobbs